[DirectX] Set whole-module flags prior to evaluating per-function flags

llvm/lib/Target/DirectX/DXILShaderFlags.cpp

@@ -32,7 +32,7 @@
 using namespace llvm;
 using namespace llvm::dxil;
 
-static bool hasUAVsAtEveryStage(DXILResourceMap &DRM,
+static bool hasUAVsAtEveryStage(const DXILResourceMap &DRM,
                                 const ModuleMetadataInfo &MMDI) {
   if (DRM.uavs().empty())
     return false;
@@ -143,7 +143,7 @@ void ModuleShaderFlags::updateFunctionFlags(ComputedShaderFlags &CSF,
   }
 
   if (CSF.LowPrecisionPresent) {
-    if (CanSetNativeLowPrecisionMode)
+    if (CSF.NativeLowPrecisionMode)
       CSF.NativeLowPrecision = true;
     else
       CSF.MinimumPrecision = true;
@@ -207,26 +207,73 @@ void ModuleShaderFlags::updateFunctionFlags(ComputedShaderFlags &CSF,
   }
 }
 
+/// Set shader flags that apply to all functions within the module
+ComputedShaderFlags
+ModuleShaderFlags::gatherGlobalModuleFlags(const Module &M,
+                                           const DXILResourceMap &DRM,
+                                           const ModuleMetadataInfo &MMDI) {
+
+  ComputedShaderFlags CSF;
+
+  // Set DisableOptimizations flag based on the presence of OptimizeNone
+  // attribute of entry functions.
+  if (MMDI.EntryPropertyVec.size() > 0) {
+    CSF.DisableOptimizations = MMDI.EntryPropertyVec[0].Entry->hasFnAttribute(
+        llvm::Attribute::OptimizeNone);
+    // Ensure all entry functions have the same optimization attribute
+    for (const auto &EntryFunProps : MMDI.EntryPropertyVec)
+      if (CSF.DisableOptimizations !=
+          EntryFunProps.Entry->hasFnAttribute(llvm::Attribute::OptimizeNone))
+        EntryFunProps.Entry->getContext().diagnose(DiagnosticInfoUnsupported(
+            *(EntryFunProps.Entry), "Inconsistent optnone attribute "));
+  }
+
+  CSF.UAVsAtEveryStage = hasUAVsAtEveryStage(DRM, MMDI);
+
+  // Set the Max64UAVs flag if the number of UAVs is > 8
+  uint32_t NumUAVs = 0;
+  for (auto &UAV : DRM.uavs())
+    if (MMDI.ValidatorVersion < VersionTuple(1, 6))
+      NumUAVs++;
+    else // MMDI.ValidatorVersion >= VersionTuple(1, 6)
+      NumUAVs += UAV.getBinding().Size;
+  if (NumUAVs > 8)
+    CSF.Max64UAVs = true;
+
+  // Set the module flag that enables native low-precision execution mode.
+  // NativeLowPrecisionMode can only be set when the command line option
+  // -enable-16bit-types is provided. This is indicated by the dx.nativelowprec
+  // module flag being set
+  // This flag is needed even if the module does not use 16-bit types because a
+  // corresponding debug module may include 16-bit types, and tools that use the
+  // debug module may expect it to have the same flags as the original
+  if (auto *NativeLowPrec = mdconst::extract_or_null<ConstantInt>(
+          M.getModuleFlag("dx.nativelowprec")))
+    if (MMDI.ShaderModelVersion >= VersionTuple(6, 2))
+      CSF.NativeLowPrecisionMode = NativeLowPrec->getValue().getBoolValue();
+
+  // Set ResMayNotAlias to true if DXIL validator version < 1.8 and there
+  // are UAVs present globally.
+  if (CanSetResMayNotAlias && MMDI.ValidatorVersion < VersionTuple(1, 8))
+    CSF.ResMayNotAlias = !DRM.uavs().empty();
+
+  return CSF;
+}
+
 /// Construct ModuleShaderFlags for module Module M
 void ModuleShaderFlags::initialize(Module &M, DXILResourceTypeMap &DRTM,
-                                   DXILResourceMap &DRM,
+                                   const DXILResourceMap &DRM,
                                    const ModuleMetadataInfo &MMDI) {
 
   CanSetResMayNotAlias = MMDI.DXILVersion >= VersionTuple(1, 7);
   // The command line option -res-may-alias will set the dx.resmayalias module
   // flag to 1, thereby disabling the ability to set the ResMayNotAlias flag
   if (auto *ResMayAlias = mdconst::extract_or_null<ConstantInt>(
           M.getModuleFlag("dx.resmayalias")))
-    CanSetResMayNotAlias = !ResMayAlias->getValue().getBoolValue();
+    if (ResMayAlias->getValue().getBoolValue())
+      CanSetResMayNotAlias = false;
 
-  // NativeLowPrecisionMode can only be set when the command line option
-  // -enable-16bit-types is provided. This is indicated by the dx.nativelowprec
-  // module flag being set
-  CanSetNativeLowPrecisionMode = false;
-  if (auto *NativeLowPrec = mdconst::extract_or_null<ConstantInt>(
-          M.getModuleFlag("dx.nativelowprec")))
-    if (MMDI.ShaderModelVersion >= VersionTuple(6, 2))
-      CanSetNativeLowPrecisionMode = NativeLowPrec->getValue().getBoolValue();
+  ComputedShaderFlags GlobalSFMask = gatherGlobalModuleFlags(M, DRM, MMDI);
 
   CallGraph CG(M);
 
@@ -252,7 +299,7 @@ void ModuleShaderFlags::initialize(Module &M, DXILResourceTypeMap &DRTM,
         continue;
       }
 
-      ComputedShaderFlags CSF;
+      ComputedShaderFlags CSF = GlobalSFMask;
       for (const auto &BB : *F)
         for (const auto &I : BB)
           updateFunctionFlags(CSF, I, DRTM, MMDI);
@@ -273,43 +320,6 @@ void ModuleShaderFlags::initialize(Module &M, DXILResourceTypeMap &DRTM,
       // Merge SCCSF with that of F
       FunctionFlags[F].merge(SCCSF);
   }
-
-  // Set DisableOptimizations flag based on the presence of OptimizeNone
-  // attribute of entry functions.
-  if (MMDI.EntryPropertyVec.size() > 0) {
-    CombinedSFMask.DisableOptimizations =
-        MMDI.EntryPropertyVec[0].Entry->hasFnAttribute(
-            llvm::Attribute::OptimizeNone);
-    // Ensure all entry functions have the same optimization attribute
-    for (const auto &EntryFunProps : MMDI.EntryPropertyVec)
-      if (CombinedSFMask.DisableOptimizations !=
-          EntryFunProps.Entry->hasFnAttribute(llvm::Attribute::OptimizeNone))
-        EntryFunProps.Entry->getContext().diagnose(DiagnosticInfoUnsupported(
-            *(EntryFunProps.Entry), "Inconsistent optnone attribute "));
-  }
-
-  // Set ResMayNotAlias to true if DXIL validator version < 1.8 and there
-  // are UAVs present globally.
-  if (CanSetResMayNotAlias && MMDI.ValidatorVersion < VersionTuple(1, 8))
-    CombinedSFMask.ResMayNotAlias = !DRM.uavs().empty();
-
-  // Set the module flag that enables native low-precision execution mode. This
-  // is needed even if the module does not use 16-bit types because a
-  // corresponding debug module may include 16-bit types, and tools that use the
-  // debug module may expect it to have the same flags as the original
-  CombinedSFMask.NativeLowPrecisionMode = CanSetNativeLowPrecisionMode;
-
-  // Set the Max64UAVs flag if the number of UAVs is > 8
-  uint32_t NumUAVs = 0;
-  for (auto &UAV : DRM.uavs())
-    if (MMDI.ValidatorVersion < VersionTuple(1, 6))
-      NumUAVs++;
-    else // MMDI.ValidatorVersion >= VersionTuple(1, 6)
-      NumUAVs += UAV.getBinding().Size;
-  if (NumUAVs > 8)
-    CombinedSFMask.Max64UAVs = true;
-
-  CombinedSFMask.UAVsAtEveryStage = hasUAVsAtEveryStage(DRM, MMDI);
 }
 
 void ComputedShaderFlags::print(raw_ostream &OS) const {

llvm/lib/Target/DirectX/DXILShaderFlags.h

@@ -85,22 +85,25 @@ struct ComputedShaderFlags {
 };
 
 struct ModuleShaderFlags {
-  void initialize(Module &, DXILResourceTypeMap &DRTM, DXILResourceMap &DRM,
-                  const ModuleMetadataInfo &MMDI);
+  void initialize(Module &, DXILResourceTypeMap &DRTM,
+                  const DXILResourceMap &DRM, const ModuleMetadataInfo &MMDI);
   const ComputedShaderFlags &getFunctionFlags(const Function *) const;
   const ComputedShaderFlags &getCombinedFlags() const { return CombinedSFMask; }
 
 private:
-  // Booleans set by module flags
-  bool CanSetResMayNotAlias;         // dx.resmayalias
-  bool CanSetNativeLowPrecisionMode; // dx.nativelowprec
+  // This boolean is inversely set by the LLVM module flag dx.resmayalias to
+  // determine whether or not the ResMayNotAlias DXIL module flag can be set
+  bool CanSetResMayNotAlias;
 
   /// Map of Function-Shader Flag Mask pairs representing properties of each of
   /// the functions in the module. Shader Flags of each function represent both
   /// module-level and function-level flags
   DenseMap<const Function *, ComputedShaderFlags> FunctionFlags;
   /// Combined Shader Flag Mask of all functions of the module
   ComputedShaderFlags CombinedSFMask{};
+  ComputedShaderFlags gatherGlobalModuleFlags(const Module &M,
+                                              const DXILResourceMap &,
+                                              const ModuleMetadataInfo &);
   void updateFunctionFlags(ComputedShaderFlags &, const Instruction &,
                            DXILResourceTypeMap &, const ModuleMetadataInfo &);
 };

llvm/test/CodeGen/DirectX/ShaderFlags/disable-opt-cs.ll

@@ -8,7 +8,7 @@
 ; CHECK-NEXT: ;       Disable shader optimizations
 
 ; CHECK: ; Shader Flags for Module Functions
-; CHECK: ; Function main : 0x00000000
+; CHECK: ; Function main : 0x00000001
 ; The test source in this file generated from the following command:
 ; clang -cc1 -triple dxil-pc-shadermodel6.0-compute -x hlsl -emit-llvm -O0 -o - <<EOF
 ; [numthreads(1,1,1)]

llvm/test/CodeGen/DirectX/ShaderFlags/disable-opt-lib.ll

@@ -8,7 +8,7 @@
 ; CHECK-NEXT: ;       Disable shader optimizations
 
 ; CHECK: ; Shader Flags for Module Functions
-; CHECK: ; Function main : 0x00000000
+; CHECK: ; Function main : 0x00000001
 ; The test source in this file generated from the following command:
 ; clang -cc1 -triple dxil-pc-shadermodel6.3-library -x hlsl -emit-llvm -O0 -o - <<EOF
 

llvm/test/CodeGen/DirectX/ShaderFlags/max-64-uavs-array-valver1.5.ll

@@ -13,7 +13,7 @@ target triple = "dxil-pc-shadermodel6.7-library"
 ; CHECK:    UAVs at every shader stage
 ; CHECK-NOT:    64 UAV slots
 
-; CHECK: Function test : 0x00000000
+; CHECK: Function test : 0x00010000
 define void @test() "hlsl.export" {
   ; RWBuffer<float> Buf : register(u0, space0)
   %buf0 = call target("dx.TypedBuffer", float, 1, 0, 1)

llvm/test/CodeGen/DirectX/ShaderFlags/max-64-uavs-array-valver1.6.ll

@@ -14,7 +14,7 @@ target triple = "dxil-pc-shadermodel6.7-library"
 ; CHECK:        UAVs at every shader stage
 ; CHECK:        64 UAV slots
 
-; CHECK: Function test : 0x00000000
+; CHECK: Function test : 0x00018000
 define void @test() "hlsl.export" {
   ; RWBuffer<float> Buf : register(u0, space0)
   %buf0 = call target("dx.TypedBuffer", float, 1, 0, 1)

llvm/test/CodeGen/DirectX/ShaderFlags/max-64-uavs.ll

@@ -13,7 +13,7 @@ target triple = "dxil-pc-shadermodel6.7-library"
 ; CHECK:       64 UAV slots
 
 ; Note: 64 UAV slots does not get set per-function
-; CHECK: Function test : 0x00000000
+; CHECK: Function test : 0x00008000
 define void @test() "hlsl.export" {
   ; RWBuffer<float> Buf : register(u0, space0)
   %buf0 = call target("dx.TypedBuffer", float, 1, 0, 1)

llvm/test/CodeGen/DirectX/ShaderFlags/res-may-not-alias-sm6.7.ll

@@ -19,7 +19,7 @@ target triple = "dxil-pc-shadermodel6.7-library"
 ; CHECK:       Any UAV may not alias any other UAV
 ;
 
-; CHECK: Function loadUAV : 0x00000000
+; CHECK: Function loadUAV : 0x200010000
 define float @loadUAV() #0 {
   %res = call target("dx.TypedBuffer", float, 1, 0, 0)
       @llvm.dx.resource.handlefrombinding(i32 0, i32 0, i32 1, i32 0, i1 false)
@@ -29,7 +29,7 @@ define float @loadUAV() #0 {
   ret float %val
 }
 
-; CHECK: Function loadSRV : 0x00000010
+; CHECK: Function loadSRV : 0x200010010
 define float @loadSRV() #0 {
   %res = tail call target("dx.RawBuffer", float, 0, 0)
       @llvm.dx.resource.handlefrombinding(i32 0, i32 0, i32 1, i32 0, i1 false)

llvm/test/CodeGen/DirectX/ShaderFlags/uavs-at-every-stage-lib-valver1.7.ll

@@ -12,7 +12,7 @@ target triple = "dxil-pc-shadermodel6.5-library"
 ; CHECK: Note: shader requires additional functionality:
 ; CHECK:        UAVs at every shader stage
 
-; CHECK: Function test : 0x00000000
+; CHECK: Function test : 0x00010000
 define void @test() "hlsl.export" {
   ; RWBuffer<float> Buf : register(u0, space0)
   %buf0 = call target("dx.TypedBuffer", float, 1, 0, 1)

llvm/test/CodeGen/DirectX/ShaderFlags/uavs-at-every-stage-vs.ll

@@ -13,7 +13,7 @@ target triple = "dxil-pc-shadermodel6.5-vertex"
 ; CHECK: Note: shader requires additional functionality:
 ; CHECK:        UAVs at every shader stage
 
-; CHECK: Function VSMain : 0x00000000
+; CHECK: Function VSMain : 0x00010000
 define void @VSMain() {
   ; RWBuffer<float> Buf : register(u0, space0)
   %buf0 = call target("dx.TypedBuffer", float, 1, 0, 1)

llvm/test/CodeGen/DirectX/ShaderFlags/use-native-low-precision-1.ll

@@ -18,19 +18,19 @@ target triple = "dxil-pc-shadermodel6.7-library"
 ;CHECK-NEXT: ;
 ;CHECK-NEXT: ; Shader Flags for Module Functions
 
-;CHECK-LABEL: ; Function add_i16 : 0x00000020
+;CHECK-LABEL: ; Function add_i16 : 0x00800020
 define i16 @add_i16(i16 %a, i16 %b) "hlsl.export" {
   %sum = add i16 %a, %b
   ret i16 %sum
 }
 
-;CHECK-LABEL: ; Function add_i32 : 0x00000000
+;CHECK-LABEL: ; Function add_i32 : 0x00800000
 define i32 @add_i32(i32 %a, i32 %b) "hlsl.export" {
   %sum = add i32 %a, %b
   ret i32 %sum
 }
 
-;CHECK-LABEL: ; Function add_half : 0x00000020
+;CHECK-LABEL: ; Function add_half : 0x00800020
 define half @add_half(half %a, half %b) "hlsl.export" {
   %sum = fadd half %a, %b
   ret half %sum